Clustering analysis

all particles

all particles - descriptive stats

1: clutering only particles with volume greater than or equal to 2mm^3

2: 1. particles greater than or equal to 2 mm^3; clustering based on coordinates

based on elbow method and silhouette score two clusters were identified as the ideal K

2 sample t-test, respectively we are going to compare cluster means for anisotropy, elongation, flatness and sphericity

significance level is set to 0.05

H0 - there is no difference between means of two clusters

2. 2., particles greater than or equal to 2 mm^3; clustering based on physical features (anisotropy, elongation, flatness, sphericity)

PCA + K-means++

two components were chosen as they capture more than 80% of all variability

loadings - correlation between pca components and original features

pca scores

based on the elbow method we're choosing 3 clusters

K-means++ clustering with PCA results

table of particles with the volume greater than or equal to 2 mm^3, 2 pca component created based on physical features (anisotropy, elongation, flatness, sphericity) and based on the 3 clusters were designated by using K means++

mean values for different clusters

three clusters: 0: 19 particles, 1: 25 particles, 2: 25 particles

as we can see above there is slightly distinguished yellow cluster number 2, typical of high value of anisotropy, mean (0,89)

2.3. clustering - particles greater than or equal to 1 mm^3

trying to gain more insight into these clusters we can try to include all particles with volume greater than or equal to 2 mm^3

all particles greater than or equal to 1 mm^3 clustering using PCA on their physical attributes

again choosing 3 clusters

as we can see above if considering particles greater than or equal to 1 mm^3, no significant cluster appear

3. Slicing venus based on vertical coordinates and exploring each slice

total height of the figurine is (cca 114.099 mm), figure was dived into 6 even parts (cca 19.016 mm) along vertical axis, using BarycenterZ coordinates

intervals for all designated slices

very top (head) - part 1

part 2 upper torso

part 3 - belly and breasts

part 4 - bottom part of belly and bottom

part 5: upper part of legs

lower part of legs, part 6

table = ratio of particles with the volume greater than or equal to 2 mm^3 to all particles in a particular part of the venus

As we can see above even dividing venus into 6 even vertical parts using BaryCenterZ coordinates, does not bring any significant results regarding distribution larger particles greater than or equal to 2 mm^3

mean values of physical features and volume fo each slice seperately (for all particles)

mean values of physical features and volume fo each slice seperately (only for particles with volume greater than or equal to 2 mm^3)

venus slices highlighed on 3d plot